User:Thingmaker/sandbox

Discrete Transistor Logic Circuits
The objective of this article is to present a brief history of computer logic circuits and a more detailed explanation of their design, design consideration and the conditions that dictated that design. A more specific objective is to explain the Discrete Transistor Logic Circuits to show how and why they differ from Twenty First Century Integrated Circuits.

Logic circuits are the building blocks of computers and computer like functions. Over the years they have been implemented in a variety of ways using a variety of components and based on the priorities of the time. The various implementations have names like DTL, RTL, CML, ECL, TTL and a few more. Nearly all have the same common functions that are basically AND, OR and INVERT or some combination of these.

Looking back at these circuits from the twenty-first century, most of these seem to be the result of some strange choices or possibly poor engineering… or maybe not.

What is a Logic Circuit?
AND, OR and INVERT are the basic building blocks of computers and computer like electronics. The circuits that provide these functions are called Logic Circuits because AND, OR and INVERT circuits perform the logical function their names imply. In the simplest form these circuits use binary or two state signals that are normally called “1” and “0”. The labeling is arbitrary and is significant only as the Logic Designer chooses. In the circuit itself the two levels are typically two approximate voltage levels such as 0 and +5 volts or possibly 0 and -6 volts. Some used + or – 0.4 volts for one set of levels and -6 volts + or – 0.4volts. These voltages have tolerances and are usually chosen by the Logic Gate circuit Designer. To the Logic Designer using them they are simply “1” and “0”.

AND, OR Example
As a simplistic example we might consider the logic to fire a Nike Ajax guided missile. We might have a signal, A that is “1” if the launcher is raised and “0” if it is down. B is “1” if that missile is selected out of the choices of twelve, and C is “1” if the Fire switch is switched “ON”. Using a Three Way AND logic gate we might generate the Fire Command “F”. If A AND B AND C are all “1”s then F will be “1” (AxBxC=F)  and the missile’s booster will receive 120 volts and fire. If A OR B OR C are “0” then “F” will be “0” and the missile will not fire. This actual function was performed using relay logic circuits in the mid 1950’s.

In 1958 The World’s Largest Transistorized Computer used 20,000 transistors providing nearly as many AND Invert and OR Invert gates to achieve a specialized computer like function. (This computer was a government secret project that was not made public.) In the Twenty First Century a processor in a personal computer might use 2,000,000,000 transistors to provide a “modest” computer function.

Integrated Logic Circuit Implementation
In the Twenty First Century we know logic circuits from their implementation in Integrated Circuits. In the mid 1960’s Integrated Circuits or IC’s began to provide the implementation for most new logic circuits. In a single Silicon chip it was possible to implement several to as many as hundreds of logic gates. The Silicon chip was a semiconductor, diamond like, crystal wafer that could have imbedded many bipolar transistors and diodes. These transistors and diodes were ideal components for the IC. Resistors were achievable but less desired and with poorer tolerance than discrete resistors. Capacitors and inductors were extremely undesirable. Not only were the transistors and diodes the ideal components of the IC but they could be made with reasonably good tolerances and very good matching characteristics from one transistor to another on the same chip. The transistors could be turned OFF even with as much as 0.4 volts of forward bias across the base/emitter and the diode junctions had similar characteristics.

These properties led to the logic families known as TTL and ECL which use many transistors with matched parameters for each logic gate. Transistors were good and diodes were nearly as good. Resistors were tolerated as a necessary component but to be used as sparingly as possible. Capacitors must be avoided except in very small values and only if absolutely necessary. Inductors were out of the question. The schematic to the right is a simple AND Invert gate showing the extensive use of transistors in an integrated circuit

Discrete Transistor Logic Circuits
By the time Integrated Circuits became a reality the computer industry was well on its way. In April 1955 the IBM 608 computer was announced. The 608 was the world’s first all transistorized calculator to be manufactured for commercial markets. IBM System/360 was announced on April 7, 1964. These and many more computer products were designed with Discrete Transistor Logic. It was Discrete Transistor Logic that launched the computer age!

What was different about discrete transistor logic and integrated circuits? In the 1950’s and early1960’s only discrete transistors were available. Many were hand made under microscopes. There were some Silicon NPN transistors but they were expensive and too slow for most computer applications. Discrete Germanium transistors were available in both PNP and NPN with the PNP being somewhat more desirable to produce.

A typical worst case specification for a germanium discrete transistor for a switching circuit would be a minimum base to collector current gain of 20 with no limit on the maximum. The base to emitter voltage required to saturate the collector was 0.4 volts and 0.1 volts reverse bias was specified to guarantee the transistor was off. The leakage current at 85º C could be 100 micro amps or 0.1 ma. The collector to emitter saturation voltage was a maximum of 0.2 volts and a minimum of 0 volts.

Each transistor suitable for logic circuit design cost around $5 or more. Diodes were also germanium and cost 25 cents for a good specification. With a specification of 1 volt forward drop and 50 micro amps leakage most fallout from small germanium diode lines could be used at a price of 11 cents. When the usage exceeded all the fallout from around the world that price went up to 15 cents.

Half watt carbon resistors with a tolerance of +/- 5 percent cost 4.5 cents. One percent carbon resistors cost around 80 cents making them undesirable except if absolutely needed. Small capacitors in the range of 100 pico farads cost 5 cents. (Prior to 1960, in the US, pico farads were called micro-micro farads or mickey-mikes for short.)

With these components resistors and small capacitors were cheap; diodes were only a little worse. Transistors were a necessity but should be used sparingly. Also transistors did not have matching parameters since they were hand made one at a time. This was nearly the reverse of what integrated circuits would bring in the years to come.

One simple transistor logic design might be built millions of times with tens of thousands used in a single computer. Reliability requirements were extreme. The design theory taught even in the finest technology universities in the world would not meet such demands.

IBM 608 Logic Circuits
The 608 cards used in the First all-transistorized computer and the 1958 World’s Largest Computer were DTL circuits. They were designed using handmade germanium PNP and NPN Alloy-junction transistors. At the time there was some doubt transistors were sufficiently reliable. The circuits were packaged on large phenolic circuit boards or cards. The cards were about 1/8” thick and maybe 6” x 7” with printed circuits on the back and all components mounted on the front. It had printed circuit tabs on one edge to allow insertion into a socket. The transistors were mounted in sockets to make replacement easy. There were some double length cards that were so long they would warp and twist.

Note the symbols for the transistors. There were no standardized symbols for the transistor at that time. IBM chose a symbol that somewhat resembled its physical construction and the graphics used to teach transistor physics. These symbols assisted visualizing the flow of majority and minority carriers through the junctions and base. Since IBM never shared their designs with the outside world they continued the use of these symbols into the 1990’s and probably into the Twenty-first Century. The logic family included an AND Invert Gate using an NPN transistor and an OR Invert Gate using a PNP transistor. The inputs and outputs were all compatible with signal swings from about 0 volts to -5 volts with collector voltages clamped with diodes in the off condition. (A description of the design and functions of DTL will follow.) A single card held four Two Way gates or three Three Way gates or one Eight Way. This was a tremendous improvement over the previous Vacuum Tube generation. There were also cards with only Inverters.

The circuit family was specified to pass minimum one microsecond pulse widths. To achieve this speed it was necessary to include the Speed-up Capacitor around the input resistor to help remove the minority carriers in the base when coming out of saturation. It also helped speed up the turn on and transition times. Without the Speed-up Capacitor the speed would have been about one third as fast.

‘’’A special Trigger circuit’’’ was included that could replace simple Latch functions as well as Shift Register and Counter functions. The Trigger was edge triggered like the D Flip Flop of future Integrated Circuits. This function was unheard of in the universities. The key was the Harper Gate consisting of a diode, capacitor and resistor. In addition to simple convenience this circuit allowed a more efficient clocking system since such a function otherwise would require two latches with phased clocks.

Harper Gate
If the signal on the resistor input (Din or Not Din) was 0 volts for a minimum set-up time prior to the positive Clock transient the capacitor/diode junction would be 0 volts. The capacitor would differentiate the Clock rise time and forward bias the diode causing the transistor base voltage to reverse bias and the transistor would turn off. When that transistor turned off it would turn on the other and become latched in that state. If the Harper Gate resistor input was -5 volts prior to the Clock's positive transient the capacitor/diode junction would be -5 volts and the differentiated clock transient would only rise to 0 volts thus not forward biasing the diode and no change would occur from that gate. At the time of the positive clock pulse the resistor (D) inputs could change without changing the effect of the capacitor on the diode and transistor.

Test Stuff
User talk:Thingmaker

Does the 550 switching diodes for the clock in the Diode–transistor logic prove that the PN diode is a switch? I have the explanation of how a diode switch works. Click on this User talk:Thingmaker. The simple PN diode chip puts integrated circuits to shame. Possibly instead of looking for published documentation of “someone” calling the diode a switch one might just read p-n junction or p-n diode and see if they can find how the pn diode is a switch.Thingmaker (talk) 20:34, 27 August 2014 (UTC)

DTL

Diode logic

Diode

CMOS

Diode law

Understanding the Shockley diode equation
From the 1950’s to well into the 1960’s discrete semi-conductor transistors and diodes provided the basis to launch the transistorized computer industry. These devices were primarily implemented in germanium crystals. Some silicon devices were available but were too expensive and too slow for most computer applications.

The physics of the pn junction including the Shockley equation was taught in any transistor electronics course but contributed little to the theory of circuit design. The equation was assumed to be an idealized representation that fell far short of any practical value.

Germanium transistor base emitter junctions and diodes had forward voltage drops of a maximum of 0.25 to 0.4 volts and a minimum of zero over a usable temperature range. The diode forward voltage was often specified at 1 volt maximum. The base emitter was often specified to require 0.1 volts reverse bias to insure its turn off. The Shockley equations were lost in the specification.

Then in the early 1960’s silicon integrated circuits started to become available. In addition to some small scale integration of complete circuits the process also provided discrete chips of diodes and transistors. These devices exhibited the high forward voltage drops of silicon but there was something new. They had little current flow at voltages as high as 0.4 volts or more. A closer look resulted in the conclusion that the Shockley equation was alive and well. A close analysis of actual diodes and transistor base emitter drops showed the voltage current when plotted on a log graph was a straight line just as Shockley’s equation suggested. This was true from currents near the leakage current level to currents near the maximum ratings.

For most silicone transistor base emitter drops and diodes the modified Shockley equation below is accurate for reverse voltages less than any effect of reverse break down to forward currents below the effect of the resistance of the P and N regions not near the junciton:


 * $$Id=I_\mathrm{S} \left( e^{V_\mathrm{D}/(n 26mv)}-1 \right),\,$$

The 26mv assumes that the thermal voltage VT is approximately 25.85 mV at 300°K, a temperature close to "room temperature". Any deviation from this value due to temperature is absorbed by the value n and the temperature effect on IS.

This equation can better be visualized by the images below. The images were calculated from the above equation but scaled (IS scaled) to an actual silicon NPN transistor base emitter specification for n=1 and an actual silicon diode specification for n=2. The variations in the specification will be added in the following discussion.

The fallacy of the knee


It is often taught and assumed that PN diodes (junctions) have a knee as seen on a linear plot of its voltage current characteristic. The top plot shows two plots one in blue the other in green of the same junction. The blue plot appears to have a knee at about 0.65 volts. That plot is accurate to the current scale on the left. The green plot appears to have a knee at about 0.55 volts. This plot uses a current of 1/100 so the point where the blue plot has a value of 10 ma the green has only 0.1 ma. Both are of the same PN junction. Only the current scale has been changed. The shift of the knee between the two is an illusion. There is no true knee or discontinuity. If you try to zoom in on the knee by decreasing the current scale the knee will shift left and keep the same shape. This will be further demonstrated in the discussion of the lower plot.

The effect of n on the linear plot
Both the blue and green plot on the upper graph indicates the curve with n=1. The red plot represents n=2. The curve softens or bends out with a higher n. For n=2 the curve starts to rise at a lower voltage but rises slower as the plot shows

The effect of T on the linear plot
The effect of temperature might be shown by the green curve if it is assumed that its scale is the same as the blue scale. The shift from the blue to the green plot would be a result of the green plot representing an increase of about 50°C as compared to the blue plot.

Log plot of the Shockley equation
The lower graph on the right shows the Shockley equation with n included plotted with the current on a log scale. The exponential nature of the equation makes a log plot useful. Except for very low or very high currents it is always a straight line when the current is plotted on a log scale and the voltage on a linear scale. On this plot there is no suggestion of a knee. The straight line will hold for currents down till it approaches the leakage current IS which is extremely small for a silicon junction. The red line that drops down from the red straight line plot at the bottom illustrates this effect. This is caused by the -1 in the equation. When the current from the exponential equals IS they cancel each other out resulting in a premature value of zero current. This effect is negligible for most practical concerns.

The straight line also deviates at high currents. This is due to the resistance of the silicon P and N regions that are not part of the junction depletion region. This is shown by the red line at the top that bends to the right from the straight line. At very high currents the slop of the linear plot would become that of a simple resistor causing an increase in voltage as compared to the exponential plot. This can be a significant deviation from the ideal Shockley equation but only at high currents. This effect tends to scale with the intended current capacity of the particular diode type. A low current diode or transistor might stay linear on the log scale till it starts to deviate about 1/10th of the maximum intended current such as 1ma for a 10ma device. Higher current diodes are more likely to show this effect sooner as compared to their maximum current range.

Understanding the PN diode junction according to Shockley can transform the diode from a simple rectifier or nonlinear resistance to a powerful circuit tool to perform log and anti-log functions for multiplication of voltages or a sine wave generator or used in feedback to form an ideal limiter for small voltages and more.

Diode logic
Diode logic (DL), or diode-resistor logic, is the construction of Boolean logic gates from diodes.

While diode logic has the advantage of simplicity, the lack of an amplifying stage in each gate limits its application. Not all logical functions can be implemented in diode logic alone; only the non-inverting logical AND and logical OR functions can be realized by diode gates. If several diode logic gates are cascaded, the voltage levels at each stage are significantly changed, so one-stage applications are normally used. In some controlled circumstances a combined design can achieve two levels feeding a single inverter amplifier. This would result in a total function of AND OR Invert or OR AND Invert. The most common use for diode logic is in DTL (Diode Transistor Logic) circuits that include an inverter for power gain and signal restoration.

Diode logic gate versions
In logic gates, logical functions are performed by parallel or series connected switches (such as relay contacts or insulated gate FET’s like CMOS) controlled by logical inputs or parallel resistors or diodes which are passive components. Diode logic is implemented by diodes which exhibit low impedance when forward biased and a very high impedance when reverse biased. There are two kinds of diode logic gates - OR and AND. It is not possible to construct NOT (Invert) diode gates because the NOT or Invert function requires an active component such as a transistor.

Simplifying assumptions
Since the middle of the twentieth century the most commonly used diodes for diode logic have been solid state p-n diodes fabricated with silicon or germanium crystals. These diodes exhibit a low forward impedance allowing it to conduct a significant current with low voltage drops, typically around a half volt and the reverse direction is a very high impedance exhibiting a near open circuit. To simplify this discussion it will assume a perfect diode that conducts forward current with a zero voltage drop and no current when reverse biased to high voltages.

Logic design using AND’s and OR’s assume binary or two levels of signal voltages that are labeled “1” or “0”. For positive logic the 1 represents the most positive level and 0 for the most negative level. For positive logic in this discussion 1 represents +6 volts and 0 represents ground or 0 volts.

OR logic gate
The image to the right shows a typical diode OR circuit. The diode symbol is an arrow showing the forward low impedance direction of current flow. All diodes have inputs on their anodes and their cathodes are connected together to drive the output. RD is connected from the output to -6 volts to provide bias current for the diodes and output drive current even down to 0 volts.

If all inputs A and B and C are logic level 0 RD will pull the output negative till the diodes clamp the output to 0 volts or a logical 0 output level. If input A switches to a logical 1 level it will pull the output voltage with it providing an output of 1. In this case diodes B and C would be reverse biased with +6 volts on their cathodes and 0 volts on their anodes. If input A OR B OR C is 1 the output will be 1. Only if all inputs, A and B and C are 0 will the output be zero. This is the definition of a logic OR. The truth table on the right of the image shows the output for all combinations of inputs.

This can be written as:
 * A OR B OR C = OUTPUT
 * or
 * A+B+C=OUTPUT

In Boolean algebra the plus sign (+) is used to denote OR.

It should be noted that RD can return to any voltage that is not more positive than the logic level 0. If RD is connected to 0 volts it will have no drive current available to drive the next circuit. All signal levels, the value of RD and its return voltage are options chosen by the circuit designer (not the using logic designer) to meet the design requirements.

AND logic gate


The diode AND is basically the same as the OR except it is turned upside down. The diodes are reversed so that the cathodes are connected to the inputs and the anodes are connected together to provide the output. RD is connected to +12 volts to provide the forward bias current for the diodes and current for output drive.

If all inputs A AND B AND C are logic level 1 RD will pull the output positive till the diodes clamp the output to +6 volts or a logical 1 output level. If input A switches to a logical 0 level it will pull the output voltage with it providing an output of 0. In this case diodes B and C would be reverse biased with 0 volts on their anodes and +6 volts on their cathodes. If input A or B or C is 0 the output will be 0. Only if all inputs, A AND B AND C are 1 will the output be 1. This is the definition of a logic AND. The truth table on the right of the image shows the output for all combinations of inputs.

This can be written as:
 * A AND B AND C = OUTPUT
 * or
 * AxBxC=OUTPUT

In Boolean algebra any conventional multiply sympology denotes AND.

Similar to the diode OR, RD can return to any voltage that is not more negative than the logic level 1. If RD is connected to a voltage equal to the 1 level it will have no drive current available to drive the next circuit. All signal levels, the value of RD and its return voltage are options chosen by the circuit designer (not the using logic designer) to meet the design requirements.

Negative logic
The assignment of 1 and 0 to the positive and negative signal levels respectively is an option of the logic designer using the AND or OR circuits. With this assignment it assumes that the logic is positive. It is just as likely that the assignment might be the reversed where 1 is the negative voltage and 0 is the positive voltage. This would be negative logic. Switching between positive and negative logic is commonly used to achieve a more efficient logic design.

In Boolean algebra it is recognized that a positive logic OR is a negative logic AND. Similarly a positive logic AND is a negative logic OR.

This relationship can easily be recognized by reading the above description of their operation. In the OR it stated, “Only if all inputs, A and B and C are 0 will the output be 0.” In negative logic each node at the lower voltage would become a logic 1,  making the statement, “Only if all inputs, A AND B AND C are 1 will the output be 1.”  That is the definition of an AND function.

Similarly for the AND it was stated, “If input A or B or C is 0 the output will be 0.” In negative logic each node at the lower voltage would become a logic 1, making the statement, “If input A OR B OR C is 1 the output will be 1.”  That is the definition of an OR function.

DELETED BY Wt A logic designer does not need an inverter to choose to use negative logic. He can name and change the logic levels to suite his needs. Today with logic families providing every logic function anyone would need this is not so important. There was a whole generation of logic designers using discrete transistor circuits into the first half of the 1960’s that did their job using only AND Inverts or only OR Inverts. It was important that the AND Invert was also a negative OR Invert or they would not have had an OR function. The logic designer switched from positive to negative logic and renamed the function as it was needed.

The logical function of any arrangement of diodes can only be established if the representation of logic states by voltage levels is known.

Diode logic with real diodes


The above descriptions assumed an ideal diode with zero resistance in the forward direction and infinite resistance in the reverse direction. The circuit designer must concern himself with real diodes. The articles p-n diode and a less detailed article p-n junction describe the physics of the PN diode. After all the discussion of electrons, holes, majority and minority carriers etc. each come down to an equation that most directly relates to the circuit designer. The real PN diode actually has a voltage current characteristic similar to the curve on the right. A more specific definition can be found in the Shockley diode equation. The designer of a reliable diode logic circuit is usually limited to what the diode specification provides which is often less than the equation suggests. Typically the specification will primarily provide a maximum forward voltage drop at one or more forward currents and a reverse leakage current. It will also provide a maximum reverse voltage limited by zener or avalanche breakdown. Typical worst case specifications are shown below for both germanium and silicon PN diodes.

Germanium diode:
 * Max forward voltage at 10 ma = 1 volt @ 0 to 85 °C
 * Max reverse leakage current at 15 volts = 100 microamps @ 85 °C

Silicon diode:
 * Max forward voltage at 10 ma = 1 volt @ 0 to 125 °C
 * Max reverse leakage current at 15 volts = 1 microamps @ 85 °C

Effects of component manufacturing variations and temperature are usually included in these specifications.

More realistically the germanium forward voltage might be 0.25 to 0.4 volts but this is often not specified. The silicon leakage current might be much lower possibly 1 to 100 nanoamps.

PN diodes also have transient behaviors that might be of concern with the design. The capacitance of a PN diode between anode and cathode is inversely proportional to the reverse voltage, growing toward infinity as it approaches zero junction volts where the forward current would approach infinity. There is also a recovery concern where the current will not decrease immediately when it is switched from forward bias to reverse bias. In the case of the diode OR if two or more of the inputs are at the 1 level and one switches to 0 it will cause a glitch or increase in current in the diodes that remain at 1. This can cause a short term dip in the output voltage. In practice if the diode logic gate drives a transistor inverter, as it usually does, and the diode and transistor are of similar construction the transistor will have a similar base collector capacitance that is amplified by the transistor gain so that it will be too slow to pass the glitch. Only when the diode is of a much slower construction will it become any concern at all. In one unusual design small selenium diode discs were used with germanium transistors. The recovery time of the very slow selenium diodes caused a glitch on the inverter output. It was fixed by placing a selenium diode across the base emitter junction of the transistor making it “think” it was a selenium transistor (if there could ever be one).

Early diode logic with transistor inverter
In the mid 1950’s diode logic was used in the IBM 608 which was the First all-transistorized computer in the world. The image on the right shows two basic logic circuits packaged on 608 cards. They were designed using handmade germanium PNP and NPN Alloy-junction transistors. The PNP and NPN transistor transistor symbols are symbols used by IBM. These circuits were designed using all discrete components. A single card would hold four two way circuits or three three way or one eight way. All input and output signals were compatible. The circuits were capable of reliably switching pulses as narrow as one microsecond.

Old version Early diode logic with transistor inverter
In the mid 1950’s diode logic was used in the IBM 608 which was the First all-transistorized computer in the world. The image on the right shows the two basic logic circuits packaged on 608 cards. They were designed using handmade germanium PNP and NPN Alloy-junction transistors. The PNP and NPN transistor symbols are symbols used by IBM. These circuits were designed using all discrete components. A single card would hold four two way circuits or three three way or one eight way. All input and output signals were compatible. The circuits were capable of reliably switching pulses as narrow as one microsecond.

'''Dropped. No citation:''' In 1957 one circuit generating the clock for the “World’s largest transistorized computer” was used to pass a half microsecond pulse. It failed after three years but fortunately faster transistors were available to fix it.

Citation response
1. I would have no problem with removing the whole subject “Early diode logic with transistor inverter”. I had some doubts about adding it but I felt it would add some meaning of what diode logic might be used for.

The “world’s largest computer” in August 1958 was not the IBM 608. It was a special purpose product developed under the leadership of Bob O. Evans in his “Navy Department” at the IBM Glendale Lab in Endicott NY for a secret US government organization (not the navy). It used 20,000 transistors. The 608 used 3,000 transistors. Maybe if the claim of world’s largest was removed it would be more acceptable. Or just delete whatever you think. Incidentally, the claim of the 608 being the first all transistorized came from the Wikipedia article and also the IBM 608 site. I have no personal knowledge of that.

2. I removed the “citation” notices because I thought I provided them. I doubt if there is any public record of this machine since it was secret. I was one of about fifty people involved in the project. I stated my reference was a personal firsthand account. Ignoring whether my attempt for a citation was acceptable, should I have just added it and left the citation request. I thought if I removed it you would be alerted and would do as you did.

3. My diode logic article has few citations. It consists of common knowledge that I would think would not need many citations. It seems common practice to place citations that only seem to prove a point or point to an article that is too large to search for the reference. I used the p-n diode citation as you suggested. In fact it has very little of use. It does a fine job of presenting the solid state physics of a p-n junction but little that directly relates to an external understanding of the diode. I am hoping to add a graphic interpretation of Shockley diode equation that is more directly usable for a circuit designer or technician. The article has only one plot of a diode curve and that is grossly distorted.

If two people that understands the electronics of diode logic believe the previous version was more accurate than my work then it should be removed all together.Thingmaker (talk) 17:56, 22 August 2014 (UTC)